Map Learning with Uninterpreted Sensors and Effectors

By David Pierce, Benjamin Kuipers. 1995

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Reinforcement Learning

Citations in the corpus

The Arcade Learning Environment: An Evaluation Platform for General Agents (2012, background): “Similar ideas have been developed around the theme of lifelong learning : learning a reusable, highlevel understanding of the world from raw sensory data (Thrun & Mitchell, 1995; Pierce & Kuipers, 1997; Stober & Kuipers, 2008; Sutton et al., 2011).”
Online Evolution of Deep Convolutional Network for Vision-Based Reinforcement Learning (2014, background): “), is to combine action learning with an unsupervised learning (UL) preprocessor or “compressor” which provides a lower-dimensional feature vector that the agent receives as input instead of the raw observation [5, 8, 12, 17, 19–21].”
KnowRob: A knowledge processing infrastructure for cognition-enabled robots (2013, background): “In contrast to work on bottom-up learning of sensor models (Pierce and Kuipers, 1997) or of the robot’s kinematic configuration (Sturm et al.”
Multi-timescale nexting in a reinforcement learning robot (2011, methodology): “Pierce and Kuipers (1997) gathered sensorimotor experience from random motion on a simulated robot, and then constructed a low dimensional embedding of the sensors from observed correlations between sensor readings. The validity of the embedding was assessed by how well the constructed embedding mat”
Intrinsically motivated neuroevolution for vision-based reinforcement learning (2011, background): “TD, policy gradients, etc), where the agent simultaneously learns both the mapping from the compressed features to actions, and the features themselves from the observations provoked by the actions [1,4,11,13,14,16].”
Evolving deep unsupervised convolutional networks for vision-based reinforcement learning (2014, background): “), is to combine action learning with an unsupervised learning (UL) preprocessor or “compressor” which provides a lower-dimensional feature vector that the agent receives as input instead of the raw observation [5, 8, 11, 15, 17, 18, 19].”
The Spatial Semantic Hierarchy (2000, methodology): “For another example, if the agent has array-structured sensors such as a sonar ring or a visual image, qualitative properties of sensory ow during motion can distinguish turns and travels [62].”
On the Impact of Robotics in Behavioral and Cognitive Sciences: From Insect Navigation to Human Cognitive Development (2010, unknown): “Thus, these experiments have permitted to contradict the conclusions made by innatists from the observation that children learn their mother tongue so fast with only poor cues and stimuli [58], [121], [123]: they proposed that their brain had to adapt biologically to encode language specific contrai”
Evolving a Roving Eye for Go (2004, background): “Instead of hardcoding the control laws that guide the movement of the roving eye, Pierce and Kuipers [10] showed that they can be learned through experience in the world.”
Bootstrapping bilinear models of Simple Vehicles (2015, methodology): “Pierce and Kuipers describe how an agent can build successively more abstract levels of representation for its sensors and actuators that can be used for prediction and various navigation tasks (Pierce and Kuipers, 1997).”
From pixels to policies: A bootstrapping agent (2008, background): “Previous work on sensorimotor reconstruction [1], [2] provides methods for generating geometric models of sensor arrays based on data collected during undirected experience.”
A Multitask Representation Using Reusable Local Policy Templates (2012, background): “As a result, contexts can be learned from commonly occurring sets of states as the agent randomly explores worlds generated from the domain, much in the spirit of bootstrap learning (Pierce and Kuipers 1997).”
Biologically Inspired Approaches to Advanced Information Technology (2004, methodology): “Often it has the form of bands 300-1000 μm wide [7, 16, 18, 21, 24-26] and up to 3000 μm long [26].”
Horde: a scalable real-time architecture for learning knowledge from unsupervised sensorimotor interaction (2011, background): “Previous researchers who did learn from sensorimotor data include Pierce and Kuipers (1997), who learned spatial models and control laws, Oates et al. (2000), who learned clusters of robot trajectories, Yu and Ballard (2004), who learned word meanings, and Natale (2005), who learned goal-directed…”